Morning, and welcome to the second day of the JPMorgan Global TMC Conference. I'm Gokul Hariharan, Senior Tech Analyst at JPMorgan. I'm happy to host Silicon Motion here, President and CEO Wallace Kou, and CFO Jason Tsai. Thanks very much for being here. Maybe, Wallace, to set the stage, I think there was a period where you were grappling with the MaxLinear acquisition, and you have been kind of accelerating your R&D pipeline quite a bit since then in the last 12 to 18 months. Could you set the stage for what you've been working on in terms of products, in terms of a lot of the R&D re-acceleration, a lot of tape-out activity that has happened, just to start off with?
First of all, thank you for having us here today. Very happy to see the investor. I think in the last 18 to 20 months, we really have accomplished so many things internally and externally. Externally, after the MaxLinear deal broke, we really have started very intense communication with customer investors, especially customers, because we need really customers to stay. Also, we set a goal to rebuild the relationship with all the NAND makers. Today, we have the best close relationship with all NAND makers, right from Samsung, SK Hynix, Micron, Kioxia, SanDisk, even YMTC. Now, internally, we restructured our organization, focused on main product. We set a goal to grow our product line to gain market share. We added 250 R&D last year, and we continue to add about 150 R&D this year to support the demand from the customers.
From the client side, we developed a new PCIe Gen5 leading 2508 channel. As well, we just launched a 2504 four-channel controller. For mobile, we launched a new product like by UFS 4.0 and UFS 2.2. For MonTitan, we are also gaining strong demand through the AI server, AI storage. That is why we really very enjoy the heavy release opportunity and to grow with the AI. For automotive, besides Micron, which is our largest automotive partner and customer, we added SanDisk and Samsung last year. For our Ferri product ourselves, we also grow very rapidly. This is a very good timing for us. Also, we feel very good about where we are today in market position.
Got it. If we zoom in on enterprise, I think you've tried to crack into the data center enterprise market before. What is the approach this time around? How is it different? Could you also talk a little bit about how big is this addressable market? What portion of that addressable market are you kind of targeting right now with these enterprise solutions?
OK. For any controller maker, if you want to grow, eventually, you will move to enterprise, because this is a larger TAN for controller maker. We started to engage with the enterprise market 10 years ago through acquisition of Shannon Systems in Shanghai. We learned quite a lot from Shannon Systems. We engaged with Alibaba and Baidu. I think initially, the business is doing OK. We reached around $70-$80 million through the open channel architecture. However, this architecture is not really mainstream. It cannot sustain. I think later, really, we ran down the opportunity. We rebuilt a new team five years ago, a dedicated team with ASIC firmware, architecture teams, C modeling, testing. We also go through all the review with Hyperscaler, CSP, major NAND maker, and server maker through our spec.
We're aiming for the storage and high-density storage, leveraging our know-how in TLC and QLC, special QLC. Now, as you see the AI trend and also high demand for 2028 terabyte QLC SSD, that's why we are really very, very busy to work with our customers. I think officially, we announced there are two tier one customers and told about six customers we have. They're going to start to run in the second half more meaningfully in 2026 and 2027. I think through 2026, 2027, it should be around 5%-10% of our sales revenue. We also mentioned each of the tier one could be 5%-10% of sales revenue. We have pretty exciting and very positive. We give more conservative guidance. I think when our firmware is ready to run by year-end, we should have more demand customers that will win in 2026.
Understood. Just to zoom in on this a little bit more, is your opportunity mostly QLC related, or do you think that you can penetrate into more customers as well in addition to the QLC opportunity?
Today, China does not need a QLC SSD. I think China's maximum capacity is 32 TB. I think two years from now, China is going to demand QLC SSD. Majority demand for QLC SSD is really from the US. That is why we are so busy and focused on this, because there are not many 2 TB QLC available today, right? Just Kioxia and SanDisk. I think Micron 2 TB QLC is coming. YMTC 2 TB QLC is also coming. We want to capture the opportunity to be the second player to deliver 128 TB QLC SSD to the market by year-end.
This is specifically to China, right?
China is only TLC.
Only TLC, and second in the global market?
Yes.
Understood. In the last earnings call, in addition to the QLC NAND opportunity for enterprise, you also talked about the NVIDIA design win for BlueField 3, and also potentially working with some of the newer storage solution providers like Vast Data, et cetera. Could you help us understand how relevant and big is this opportunity compared to what you have already secured with the two tier one customers?
It's very exciting news for Silicon Motion. First of all, let me just say, in the past, we never tried to talk about our enterprise boot storage product. Nobody knows we own 100% of Google boot storage drives seen five years ago. We own 100%. Since the revenue and the volumes now compared to the client's need is too small, we never talked about it. NVIDIA is really the big potential of Silicon Motion. We started engaging with NVIDIA two years ago. I think in the beginning, they did not show strong interest to keep the dialogue. Starting from one year ago, it became busy. They not only issued the spec review, and we sent the product for qualification. Until probably late Q3 last year, we know that's a BlueField. I think most of us do not understand what exactly BlueField will do.
I think BlueField is a data processing unit, and really is acceleration for the networking. For this platform, it is used to expand the ecosystem for NVIDIA for storage system as storage bay. You look from Blackwell and through the InfiniBand and Ethernet connect to BlueField platform. Through BlueField, they have connected to PCIe switches. That is why it created a storage bay and the storage system as storage server. The opportunity for us is a design win in BlueField and the BlueField 3. I think through the collaboration and for final qualification stage, now we realize there are two suppliers. The other supplier also uses Silicon Motion controller and firmware with different NAND. Technically speaking, we own 100% of the PL3 boot storage. That is a new platform. Through the ecosystem, now we know the VAST and several others.
I think they created NVIDIA selling the BlueField platform, and they use for their storage system. There is opportunity for us not only in the BlueField boot storage, but also in the data storage with 128 TB QLC SSD. That is why we are very busy trying to the same solution can send to other ecosystem providers, such as DDN, such as NetApp, Dell, and others.
When do you think this starts to kind of kick into some revenues?
We think the BlueField 3 ramping will be late Q3 and Q4 time frame this year.
This is starting with boot storage, right? I think. And then eventually the high-capacity storage?
High-capacity coming also is similarly late this year or early next year. I think the Solidigm only company provides a 122 TB QLC SSD solution today.
Understood. Very clear. All this leads to pretty strong momentum on the enterprise side. Is there any gating factor you see or any challenges you see in terms of ramping up these customers? Like you said, even one of these tier one customers could help you reach your 5%-10% target. You issued the target even before you secured this NVIDIA win. I just wanted to understand how should we think about this in the next 18 months?
What we actually need is really the R&D resources, because each of our end customer demands is very heavy. We just do not have enough resources to meet the customer demand. You probably wonder why Silicon Motion MonTitan gets such attention and strong demand, because our pattern performance shaping is ideally fitting AI ecosystem. It looks at AI data pipeline found in ingest to transformation preparation to tokenize to training. During the training, you may see the training fail, and you need to checkpoint and backup, also inference. They all need quite a lot of data, but they need different bandwidths and peak. Now, how you can really achieve the performance efficiency, data efficiency is very critical, important for the AI ecosystem. That is why our MonTitan platform is very, very in strong demand, very unique position for AI ecosystem. We just need to keep execution.
We have a very flexible business model for OEM. Some is really using RSDK to develop firmware themselves. Some is joint development. For tier two, tier three, really our turnkey solution. We have to make sure the turnkey solution is ready, and we start to kick in revenue in the second half. Next year, we're going to see the meaningful ramp next year.
Understood. How was the NAND maker interest for your solution? Because historically, they have always stuck to their own solutions, right?
I can only say there's a couple waiting. Just NAND maker, they all have their own solution, but also have a limitation. Some cannot provide the high-density 128 TB solution. That's what we have value add to them. We create a certain ecosystem and try to enhance. I think we don't provide solution, only provide controller and firmware. We build ecosystem even for NVIDIA, like a VAST ecosystem.
Understood. If we switch a little bit to client SSDs now, you are pretty much the market leader there in the third-party controller segment. How do we grow from here? I think you have a few exciting product cycles coming up on PCIe Gen5. Could you talk a little bit about how we see the growth in the client segment?
We are about 30% market share today globally for client SSD for PC. As you see, we announced our PCIe Gen5 eight-channel controller 2508. Lately, you see the price relief from Micron, from Kingston, from SanDisk yesterday, and from AData. I think there will be more coming in the next few weeks. We see our eight-channel PCIe Gen5 will probably achieve minimum 50%-60% market share for high-end. Our four-channel controller PCIe Gen5 DRAM-less 2504, we are available just two months ago. We also quickly won four NAND maker, nearly all module maker design. We believe when they start to ramp by year-end, PCIe will ramp from middle of 2026. We should own about 50%. Through the mathematically, we should easily achieve 40% in the next three years, probably even more. We do not see near competition.
The good thing is the PCIe end will not adopt PCIe Gen6 until 2030, only for high-end. You can imagine the next three to four years we're in a dominant position in the market. NAND maker will never reduce the controller to meet their demand for existing standard. We're about 40% for PCIe Gen4. With 30%, I think out of PCIe Gen5, we should reach very quickly to 40% market share.
How does it impact the pricing dynamics also? I think we've been in that $4-$5 range for client SSD controllers on an average selling price. Do we see an upward move in that as well?
Let me just give you the comment. I think PCIe Gen5 eight-channel is around two to three times of a PCIe Gen4 four-channel controller. PCIe Gen4, PCIe Gen5 four-channel controller is about 20%-30% increase compared to PCIe Gen4. The gross margins are a lot better.
Understood. That's very clear. OK. I think maybe we move to mobile. Could you talk a little bit about your eMMC business momentum? You seem to be getting a lot of market share with the module makers, especially in China as well as other markets. Maybe we start there?
We feel very good about our mobile business because market trend is favored Silicon Motion. First of all, let me talk about eMMC. eMMC in most smartphones is declining quickly. You want to know how you grow eMMC. Because eMMC declined quickly, NAND maker tried to move away from eMMC, all sorting for server by like Silicon Motion. They look for SanDisk using our eMMC for mobile phone. Samsung decided to use our eMMC for automotive. That is why we gained market share. In addition, eMMC is a JEDEC standard. They expand beyond just smartphone, like automotive, smart TV, set-top box, and all the IoT devices. Lately, we see the AI glasses become very popular and should reach about 20 million units this year and grow much faster next year. They all use eMMC. The total trend for eMMC is around 800 million units.
We're just about 25%. We see last year. We see we're going to grow quickly this year and beyond. For mobile phone, I think in the past, all smartphone maker in mainstream and value line model, they try to use EMCP/UMCP, which means it's a controller with a NAND and mobile DRAM together in one single PGA package, because it saves the footprint and also secures supply. From early last year, China CXMT introduced LPDDR4, which is a high volume, much lower price that changed the momentum and the dialogue for the business. I think smartphone maker tried to use discrete mobile DRAM and use a standard eMMC or UFS for value line. That's why we see the momentum shift. All the module makers, they are ready to play. They increase the market share. That's why we benefit.
To say that, don't make a mistake, we also aggressively engage with NAND maker for UFS for mainstream high-end. This is our value add. We see the all-see scenario in market trend favor Silicon Motion. We're about 20% market share today, including iPhone. I think we'll grow quickly to 30% within three years.
Could you talk a little bit more about the UFS initiative? I think you had a big customer. That customer kind of went away a little bit on their latest UFS. You have had some other wins as well, not probably as big. How should we think about the UFS progression design wins?
I think UFS 2.0 is going to play eMMC in the value line. They're going to stay for the next five years, maybe forever. UFS 4 is a high-end. A year from now, UFS 5 will become high-end. Four will become mainstream. UFS 3 is in the middle. We see opportunities for Silicon Motion. In the past, we relied on engagement with NAND maker to grow our mobile business. Now we directly engage with a smartphone maker. We enable QLC to move into mobile storage. We were very successful. Transsion is our first smartphone partner. We also start to engage with the second smartphone maker, Xiaomi, this year. I think we eventually will try to engage directly. Let's create an ecosystem. Because NAND maker likes to sell standard UFS solution. We can have a customized firmware show differentiation. We are not trying to compete with NAND maker.
It's to create a better value for the NAND maker if they want to sell the wafer with module maker. They make our business model more flexible and expand opportunity for us to grow.
Understood. It's clear. Just an extension of, I think you're using eMMC also to penetrate pretty quickly into automotive. I think you talked about revenue contribution already reaching single-digit numbers already. How should we think about this market? Is it like an extension of eMMC? Or are you seeing different characteristics coming to this market also as a lot of the storage solutions also become more complicated?
Automotive is very important for Silicon Motion to grow in the long term. We set a very good strategy, engage with NAND maker, as well as our Ferri solution, selling directly to automotive maker. As we said, Micron always our largest NAND partner and customers. We added SanDisk in late 2023. We added Samsung in mid 2024. Naturally, they all use different solutions. Some use PCIe. Some use UFS. Samsung primarily focused on eMMC, transitioning to PCIe later. Now, for Ferri, as you all know, it's challenging and difficult for a merchant company to engage with a tier one automotive customer, because we don't have NAND. We cannot guarantee longevity supply. You have to have some interesting strategy and value proposition. Let me give you an example for Toyota. Toyota has three major suppliers: Denso, Denso 10, Panasonic.
We start to engage with Denso 10. Our strategy is to try to provide TLC-based storage solution, because all Toyota at the time used MLC. And because MLC is going to come to end of life, we see it's a great opportunity to provide the solution. But it's challenging because they worry about quality, data retention, and many others. We spent two years providing all the data, worked with Denso 10 R&D, and convinced them and also provided solutions for Toyota to check. Finally, we won the design and went to mass production 2023. It's for China model. Now, later, we leveraged the design win. We engaged with Denso. We won the Denso 2026 global model for ADAS. Now, lately, we just won the Panasonic design. Start with eMMC, followed by PCIe. Now we become the major storage supplier for Toyota.
I think every major automotive, we have a different strategy approach. We are very, very confident, very grateful to the business. Besides Toyota, we have Tesla. We have BYD. We have Xiaomi 100%. We have Mercedes. We have Honda. We have General Motors, Cadillac. We have Waymo and many, many others, Renault. We see the momentum coming. This year, although the price is very bloody, we see unit will grow, especially in China market.
A couple of questions on the auto side. One, I think a lot of your design wins have happened in the last maybe 12 to 18 months. When does the revenue inflection happen? Is it more like 2027 is when we should see a beam?
I think we should start to see the reflection in 2026.
Twenty-six. And how is the mix of the business? I think some of the customers, you are applying the Ferri kind of business model. Many other customers, you are largely selling controllers. So what is the long term? Is it mostly a controller sale? Or is it also like an embedded system kind of sale?
We keep open. I think Micron is the largest automotive storage player. Samsung is second. Samsung is aiming to be number one too. We are very grateful to have both of them. We are continuing support. We believe we are going to add one more NAND maker next year. We were selling both solution as well as controller to NAND maker and module maker. We aim to be the largest player in the automotive storage market.
Great to hear that. If I may now move a little bit to the near-term dynamics. I think most of the semiconductor companies, your peers, have been pretty cautious on second half 2025. In your guidance a couple of weeks back, Wallace, you still were pretty bullish that we will exit the year with a $1 billion run rate. Let's say $240-$250 million revenue in Q4. Q4 will be some kind of $240-$250 million run rate, which is a pretty big jump from where we guided Q2 to be. Could you talk a little bit about how you factor in some of the macro as well as some of the company-specific growth drivers in enterprise, in automotive, et cetera, and also the ramp-up of PCIe Gen5?
Silicon Motion definitely feels like one of the companies that's just kind of bucking the trend in terms of how you're guiding for the second half of the year. I just wanted to get into that a bit more?
I think that nobody can predict about the geopolitical issue or tariff, what it will end to. See, for most of our U.S. customers, they all have manufacturers overseas. All have manufacturing partners. They know how to manage their shipment. Just through our design pipeline and design win, we have very high confidence to reach, to achieve, or exceed our guidance. We really see the momentum, especially mobile and client in our side. Enterprise, it's a very exciting story. As I said in the earnings call, to reach a $1 billion run rate in Q4, we do not count on MonTitan. We do not count on NVIDIA BlueField 3. This is really a momentum for 2026 and 2027 automotive. We do have several major design wins. When they mature, we will have an update. Now it's really focused execution.
I think every customer demands full support, the best team. We have around 1,500 R&D right now. By support, many, many OEM customers worldwide. We just have to keep our focus, execute our plan. We definitely will deliver the result.
Understood. Maybe bring in Jason a bit here. I think about a year back or so, you outlined a gross margin target to get back to high 40s, 50%. You're almost getting there. Operating margin is the one where we're still below where we have historically reached, like mid-20s or even higher. What needs to happen from here on? If we just execute to the pipeline that we've got, do we naturally get to that kind of operating margin? I think you've also been spending a lot of money on R&D and tape outs, necessarily so. How should we think about that over the next maybe several quarters?
Yeah. Wallace highlighted a number of new products that we've been introducing, a lot of new end markets that we've been targeting. We've been spending a fair bit of money executing to those strategies. If you take a look at over the last 18 months, we've taped out three six-nanometer controllers. What we said is for each controller, you're looking at $15 million-$20 million cost. On top of the 250 engineers that we've brought on, all this is to say that we've made a lot of investments over the last 18 months. We haven't yet really started seeing the ROI on that. That's coming this year. That'll come into 2026 and 2027 longer term.
As we begin to scale some of these newer products that are higher ASP, higher gross margin, that gets us to that gross margin level of 48%-50% that we've historically achieved. As the benefits, as we start seeing that revenue kick in from these investments in SM2504, 2508, and then our UFS 4.0 controller, we'll start seeing that operating margin leverage kick in as well. We'll continue to invest. We've got additional advanced geometry products that we are taping out this year and going into next year. There are going to be continuing investments. The plus side is that we'll finally start seeing the ROI for a lot of those investments that we had made over the last 18 months. We're confident that we can get to that 48%-50% gross margin range.
We're confident that we can achieve over the midterm back to that normalized operating margin of 25% plus.
Got it. Thanks, Jason. Maybe one question on sourcing. Given you are now taping out more and more projects in leading edge, it's expensive. And pricing is going up as well. You have a more diversified portfolio with enterprise and automotive as well. How do you think about sourcing strategy, especially on the foundry side, where you have been largely sole-sourced? Do you think that there needs to be some changes there? Or you're comfortable with the current strategy?
I think TSMC is always our major partner in the sourcing for foundry. We start to see automotive customers ask to tape out in non-China, non-Taiwan, because they worry, they have a concern. We get some projects related to Samsung. We start to use Samsung as an alternative solution. That is a very small portion. We have to start to prepare a second source. TSMC also started investing in Arizona and in the U.S., so this depends how they go and what the cost structure looks like. Diversify is the direction we need to go.
From a cost perspective, I think you're obviously ramping six-nanometer-based chips. You're also considering four-nanometer tape outs. Those are all pretty expensive compared to your current portfolio, which is like 28 or 60 nanometer. How is the appetite from the customers also to kind of absorb some of those increases? Do you feel like that is what the market needs anyway? Customers will be willing to pay for that?
You are correct. From a cost structure, six-nanometer is much more expensive than 12-nanometer and 28-nanometer, and four-nanometer even more. I give you as an example, if you do not count R&D expense, six-nanometer will cost you around $15-$20 million from tape-out, mask, IP-related. You have to be one time successful. For four-nanometer, it will cost you about $35-$40 million from the mask, all the related IP. The IP is more expensive. You do not even count the R&D cost. In other ways, we see less competition coming. From PCIe Gen5, while we are in a dominant position, we have less competitor. We have few competitors. If we go to PCIe Gen6, I think probably.
Early one.
This gives us a new opportunity. Even NAND makers, they have to rethink again. Should we really do internally because it does not make sense? Or do you stick on motion controller? That will just make a solution. That is why we see more and more outsourcing opportunities from NAND makers. That will give us a very unique position because we build out the IP. Frankly speaking, we have more demand than we can handle today. We also see some custom design requests from hyperscalers, from leading automotive customers. That is why we have to be smart to utilize our cash position next year after the MaxLinear deal reaches resolution, hopefully by year end. We might have to start some articulation related to certain researchers and also mixed signal team. We are eager to grow.
We just need more R&D resources to help us support our meet the customer demand.
Understood. We've got a couple of minutes. Any questions from the investors? OK. Wallace, in the last couple of minutes, any messages that you want to share to end the conversation?
I think Silicon Motion is a great company. We have great technology. We have many, many innovations. We are in a very good position today in the market from clients being mobile, enterprise, automotive. I think we're going to continue to grow. Hopefully, we can execute the plan and provide a good update to investors.
Thanks. Thanks, Wallace. Thanks, Jason. Thank you very much for participating in the conference. Thank you.
Thank you, Gokul.